This study investigates the microvascular involvement in diabetic retinopathy (DR) using non-parametric machine learning methods. The research analyzed optical coherence tomography angiography (OCTA) images from four groups: healthy eyes, diabetes mellitus (DM) without DR, mild DR, and moderate DR. The goal was to identify key microvascular parameters that distinguish these groups and understand their role in DR progression. The study used SHapley Additive exPlanations (SHAP) values to determine the importance of these parameters in classification tasks. Key findings include that large choriocapillaris flow deficits were most important in distinguishing healthy eyes from DM without DR, but became less significant in mild or moderate DR. The superficial microvasculature was important in distinguishing healthy from DM without DR and mild from moderate DR, but not in the DM without DR to mild DR comparison. The foveal avascular zone (FAZ) metric was less affected but showed increased involvement with worsening DR. The study highlights the complex relationship between microvascular changes and DR progression, suggesting that the choroid is affected early in diabetes, while the deep retinal vascular plexus becomes more important as the disease progresses. The findings provide insights into the microvascular involvement of DM and DR, supporting the development of early detection methods and intervention strategies. The study also emphasizes the importance of considering interactions between microvascular parameters in understanding DR progression. The research underscores the potential of machine learning techniques in classifying and understanding the progression of DR, with implications for future research and clinical practice.This study investigates the microvascular involvement in diabetic retinopathy (DR) using non-parametric machine learning methods. The research analyzed optical coherence tomography angiography (OCTA) images from four groups: healthy eyes, diabetes mellitus (DM) without DR, mild DR, and moderate DR. The goal was to identify key microvascular parameters that distinguish these groups and understand their role in DR progression. The study used SHapley Additive exPlanations (SHAP) values to determine the importance of these parameters in classification tasks. Key findings include that large choriocapillaris flow deficits were most important in distinguishing healthy eyes from DM without DR, but became less significant in mild or moderate DR. The superficial microvasculature was important in distinguishing healthy from DM without DR and mild from moderate DR, but not in the DM without DR to mild DR comparison. The foveal avascular zone (FAZ) metric was less affected but showed increased involvement with worsening DR. The study highlights the complex relationship between microvascular changes and DR progression, suggesting that the choroid is affected early in diabetes, while the deep retinal vascular plexus becomes more important as the disease progresses. The findings provide insights into the microvascular involvement of DM and DR, supporting the development of early detection methods and intervention strategies. The study also emphasizes the importance of considering interactions between microvascular parameters in understanding DR progression. The research underscores the potential of machine learning techniques in classifying and understanding the progression of DR, with implications for future research and clinical practice.